Algorithmically Created Cultural Tastes: How Recommendation Algorithms in Digital Platforms Control Individuals' Music Tastes

This research investigates the influence of recommendation algorithms on individual music preferences within digital platforms, specifically how these systems either reinforce existing tastes or introduce new genres. Using data from Spotify's API and the Million Playlist Dataset (MPD), the study analyses how algorithmic recommendations impact the variety and atypicality of music consumption. The research follows an experimental methodology, creating distinct user profiles to simulate real-world music consumption and interaction with algorithmic recommendations.

The findings suggest that, over time, recommendation algorithms generally reduce genre diversity, especially for users with initially broad tastes. These users experience a narrowing of their musical exposure, with recommendations reinforcing familiar and predictable genres. This trend is quantified by a decline in the atypicality score of their music choices. In contrast, users with narrow or specific musical preferences at the outset see some diversification of their music exposure, although this expansion is constrained by genre boundaries aligned with their initial preferences.

A co-genre network was used to measure the atypicality of music genres, providing a quantitative analysis of how algorithms shape cultural consumption patterns. The study combines sociological theories of cultural capital and cultural omnivorousness with computational social science methods, offering a comprehensive view of the role that recommendation algorithms play in mediating cultural tastes.

Overall, the research highlights the dual role of algorithms in both reinforcing existing preferences and subtly reshaping cultural consumption. These findings have significant implications for understanding how digital platforms impact cultural diversity and may contribute to the perpetuation of social inequalities. Further research is needed to explore the development of fairer algorithmic systems that enhance cultural diversity while minimising bias.